Audio & Sound
20 November 2013
Sound is the most immediate application of Sunics technology. Many audio signals are naturally periodic — every note, vowel, and instrument timbre is built from repeating waveform cycles. The cyclogram transformation maps that repetition into a 2D image, producing a visual fingerprint unique to the sound.
This is not a spectrogram. It is not an artistic interpretation. Every pixel in a cyclogram encodes the data of the original audio signal. The transformation is lossless and reversible: the image IS the sound.
Timbres — Mobile App
Timbres is a mobile application that demonstrates the cyclogram transformation on recorded audio. Load a sound, and the app calculates and displays its timbre as a cyclogram. The gallery on this site is generated from Timbres.
The app shows what “see sound” means in practice. A didgeridoo vs. a Hammond organ: two completely different textures — and two completely different cyclograms. You can see it instantly.
Scinus — Desktop Application
Scinus is a desktop app that applies the transformation to live audio input. It captures audio continuously, calculates the cyclogram, and updates the display in real time as the sound changes.
The demo below was created with Scinus. A single note is held on a Minimoog synthesiser while the filter is slowly tweaked — the cyclogram evolves visibly as the timbre changes. You can often see it before you hear it.
This is because humans are significantly better at seeing visual patterns change than hearing sounds evolve.
Applications in Audio
The implications for audio production, classification, and analysis are significant:
- Timbre detection: cyclograms allow direct visual identification of sounds, independent of pitch or volume
- Visual sound design: modifying a cyclogram image (blur, sharpen, warp) modifies the timbre in intuitively predictable ways
- Image-based sound synthesis: generating images and transforming these to sounds allows for a whole new class of sounds
- Classification: image recognition techniques can classify sounds by timbre with no need for bespoke audio algorithms
These are research directions, not finished products. Sunics is an open platform — if any of these threads interest you, we hope to hear from you.